What ASCI actually requires
The Advertising Standards Council of India (ASCI) released its Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising on 29 September 2026. The text carries a Board of Governors date of 17 September 2026, and it takes effect three months after publication. It defines synthetically generated content (SGC) as audio, visual or audio-visual material that is artificially or algorithmically created or altered so that it appears real, and that depicts a person or event in a way likely to be believed as true (ASCI guidelines text).
Labelling is mandatory only where SGC "materially influences consumer decisions" and the lack of a label would mislead. The guidelines list the mandatory cases:
- synthetic influencers and ambassadors;
- replicated real voices or faces, even with consent;
- fabricated events or settings;
- demonstrations of products that do not yet exist;
- sound effects tied to a product's core features;
- paid AI-chatbot product suggestions, which must be marked "Sponsored by [Brand]".
Routine edits such as colour correction and blemish removal need no label. Nor do decorative backgrounds, obviously fantastical effects, ad-copy drafting or accessibility captions. Fabricated testimonials, exaggerated results, non-existent locations and non-consensual deepfakes are prohibited outright, and a label does not cure them.
The coverage reaches TV, print, outdoor and online, which includes ads running on streaming and online-video services (MediaNama).
The strongest case for something stricter
The case for stricter rules deserves a fair hearing. ASCI has no legal penalty power, and a materiality test lets the advertiser decide what is "material". Synthetic persuasion is cheap to produce and hard to spot, and a label that is never required cannot protect anyone. The IT Rules route is more demanding. Under the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, notified on 10 February and in force from 20 February, synthetic content must carry prominent labels and provenance metadata. Significant social media intermediaries, those with 5 million or more Indian users, must obtain user declarations and verify them (Hogan Lovells). A regulator wary of deepfake-driven fraud could reasonably want ads held to that standard.
Why the proportionate design is better
The argument against copying the IT Rules into advertising is about what a label is for. The IT Rules ask whether content is synthetic. ASCI asks whether the synthetic element changes what a consumer believes about the product. A brand that uses AI to extend a skyline or tidy a voice-over has not misled anyone about a moisturiser or a headset. Forcing a badge on it teaches viewers to ignore badges. ASCI says it wants to avoid "consumer label fatigue", and its draft stage described the approach as one that manages "consumer outcomes rather than regulating the technology" (exchange4media).
That is also the pro-innovation position. Indian creative and advertising work increasingly relies on generative tools for localisation and production cost. A trigger tied to deception leaves those uses alone and still targets the harms: fake endorsements, doctored before-and-after results, and products shown before they exist. ASCI also kept the guidelines inside its existing Code. Prohibited conduct breaches the Code whether or not it is labelled, so the label cannot be used as a shield.
The process was also sound. ASCI published a draft in May 2026 and took comments until 13 June (ASCI draft guidelines). The final text kept the draft's risk-based logic.
Where the OTT problem sits
The unresolved issues land on streaming and other platform-delivered ads. MediaNama's reading identifies several open questions (MediaNama):
- The text does not say whether a platform-applied label satisfies the brand's ASCI duty.
- It does not say how to label ads generated inside platforms.
- It does not say whether one label covers a whole ad or each synthetic segment.
- It sets no size, duration or placement rules for labels.
Those gaps matter on a connected-TV screen or in a six-second pre-roll, where a label's size and duration decide whether anyone sees it. The guidelines allow brands to use either platform-provided labels or their own, such as "Audio/Video created using AI".
The two regimes can also diverge. A realistic AI influencer on a video platform needs a brand label under ASCI and likely a platform label under the IT Rules. A decorative AI background needs neither from the brand but may still attract a platform label if it realistically depicts a person or event. A streaming service ingesting thousands of creatives will have to run two tests on every ad.
Chatbot advertising makes the question more urgent. Razorpay announced on 5 October that it will run ChatGPT ad campaigns for nine Indian brands, using an AI agent to create and manage the campaigns (MediaNama). The "Sponsored by [Brand]" rule already covers that situation, but how it applies to agent-built campaigns is untested.
What should happen before the effective date
There are three months to close the gaps without new legislation. First, ASCI should publish a short compliance note saying that a platform label meeting the IT Rules suffices where it communicates the same fact. Second, ASCI and the platforms should agree minimum label legibility for short video and connected-TV formats, as a voluntary standard. Third, MeitY should resist folding advertising into the intermediary rules. A self-regulatory layer that targets deception is cheaper, faster to update, and less likely to chill legitimate creative work than a second statutory labelling regime.
The lack of legal penalty is a limit, not a flaw. The conduct ASCI prohibits is already unlawful under consumer-protection, copyright and IT law, so ASCI's role is to set clear norms and shame offenders, with statutory enforcement behind it where it counts. If the norms prove clear, advertisers and platforms get predictability. If they prove vague, the practical outcome will be defensive over-labelling, which defeats the purpose of the materiality test.